Browsing by Author "Uguz, B."

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    Article
    Artificial Neural Network Modeling for Multi-Parameter Performance Prediction of Electronically Commutated Fan Coils Based on Experimental Data
    (Springer Science and Business Media B.V., 2025) Uguz, B.; Çolak, A.B.; Karakoyun, Y.; Gemici, Z.; Dalkılıç, A.S.
    The main problems with the selection and operation of fan coils in air conditioning systems impact thermal comfort and energy efficiency, and research on fan coil performance at various operating points is inadequate. No research on artificial neural networks has been undertaken about a concealed ceiling-type electronically commutated motor fan coil that has been subjected to extensive experimental assessments. Four artificial neural networks were trained using 1700 test points to predict the thermal performance and capacity as a main aim. The experiments were conducted in a test apparatus designed according to related standards and an indoor air and heat exchanger fluid regime based on international test norms. The first model estimated air flowrate using six input parameters. The second one estimated air outlet temperature and total cooling capacity using five input parameters. Then, the third one estimated heat exchanger fluid side pressure loss using five input parameters. Lastly, the fourth one estimated air outlet temperature, fan power, and total cooling capacity using eight-input parameters. The Levenberg–Marquardt training algorithm was employed in the feedforward backpropagation multilayer perceptron network model comprising 10 neurons in the hidden layer. The deviation obtained for the air flowrate was − 0.255% in the first one, while the deviations obtained for the air outlet temperature and cooling capacity were − 0.195, − 0.012%, respectively, in the second one. In the third one, the fluid pressure loss exhibited a deviation of − 0.014%. In contrast, the air outlet temperature, cooling capacity, and fan power exhibited deviations of + 0.045, − 0.014, and + 0.283%, respectively, in the fourth one. This study promotes energy-efficient industries using artificial intelligence-driven performance modeling as a collaboration sample between university and industry. © Akadémiai Kiadó Zrt 2025.